From cd488022ddf13238e8f0675d25ac62cf0bbda2ea Mon Sep 17 00:00:00 2001 From: 1152450298 <1152450298@qq.com> Date: Fri, 26 Dec 2025 18:44:35 +0800 Subject: [PATCH] =?UTF-8?q?=E6=9B=B4=E6=96=B0=E6=95=B0=E5=AD=97=E4=BA=BA?= =?UTF-8?q?=E6=9D=BF=E5=9D=97?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- web/components/digital_tts_config.py | 237 +++++++ web/i18n/locales/en_US.json | 64 +- web/i18n/locales/zh_CN.json | 26 + web/pipelines/__init__.py | 1 + web/pipelines/digital_human.py | 644 ++++++++++++++++++ workflows/runninghub/digital_combination.json | 4 + workflows/runninghub/digital_image.json | 4 + 7 files changed, 966 insertions(+), 14 deletions(-) create mode 100644 web/components/digital_tts_config.py create mode 100644 web/pipelines/digital_human.py create mode 100644 workflows/runninghub/digital_combination.json create mode 100644 workflows/runninghub/digital_image.json diff --git a/web/components/digital_tts_config.py b/web/components/digital_tts_config.py new file mode 100644 index 0000000..c662390 --- /dev/null +++ b/web/components/digital_tts_config.py @@ -0,0 +1,237 @@ +# Copyright (C) 2025 AIDC-AI +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Style configuration components for web UI (middle column) +""" + +import os +from pathlib import Path + +import streamlit as st +from loguru import logger + +from web.i18n import tr, get_language +from web.utils.async_helpers import run_async +from pixelle_video.config import config_manager + + +def render_style_config(pixelle_video): + """Render style configuration section (middle column)""" + # TTS Section (moved from left column) + # ==================================================================== + with st.container(border=True): + st.markdown(f"**{tr('section.tts')}**") + + with st.expander(tr("help.feature_description"), expanded=False): + st.markdown(f"**{tr('help.what')}**") + st.markdown(tr("digital_tts.what")) + + # Get TTS config + comfyui_config = config_manager.get_comfyui_config() + tts_config = comfyui_config["tts"] + + # 只使用local模式，移除inference_mode选择 + tts_mode = "local" + + # ================================================================ + # Local Mode UI (简化版本) + # ================================================================ + # Import voice configuration + from pixelle_video.tts_voices import EDGE_TTS_VOICES, get_voice_display_name + + # Get saved voice from config + local_config = tts_config.get("local", {}) + saved_voice = local_config.get("voice", "zh-CN-YunjianNeural") + saved_speed = local_config.get("speed", 1.2) + + # Build voice options with i18n + voice_options = [] + voice_ids = [] + default_voice_index = 0 + + for idx, voice_config in enumerate(EDGE_TTS_VOICES): + voice_id = voice_config["id"] + display_name = get_voice_display_name(voice_id, tr, get_language()) + voice_options.append(display_name) + voice_ids.append(voice_id) + + # Set default index if matches saved voice + if voice_id == saved_voice: + default_voice_index = idx + + # Two-column layout: Voice | Speed + voice_col, speed_col = st.columns([1, 1]) + + with voice_col: + # Voice selector + selected_voice_display = st.selectbox( + tr("tts.voice_selector"), + voice_options, + index=default_voice_index, + key="digital_tts_local_voice" + ) + + # Get actual voice ID + selected_voice_index = voice_options.index(selected_voice_display) + selected_voice = voice_ids[selected_voice_index] + + with speed_col: + # Speed slider + tts_speed = st.slider( + tr("tts.speed"), + min_value=0.5, + max_value=2.0, + value=saved_speed, + step=0.1, + format="%.1fx", + key="digital_tts_local_speed" + ) + st.caption(tr("tts.speed_label", speed=f"{tts_speed:.1f}")) + + # ================================================================ + # TTS Preview (只支持local模式) + # ================================================================ + with st.expander(tr("tts.preview_title"), expanded=False): + # Preview text input + preview_text = st.text_input( + tr("tts.preview_text"), + value="大家好，这是一段测试语音。", + placeholder=tr("tts.preview_text_placeholder"), + key="digital_tts_preview_text" + ) + + # Preview button + if st.button(tr("tts.preview_button"), key="digital_preview_tts", use_container_width=True): + with st.spinner(tr("tts.previewing")): + try: + # Build TTS params for local mode + tts_params = { + "text": preview_text, + "inference_mode": "local", + "voice": selected_voice, + "speed": tts_speed + } + + audio_path = run_async(pixelle_video.tts(**tts_params)) + + # Play the audio + if audio_path: + st.success(tr("tts.preview_success")) + if audio_path.startswith('http'): + st.audio(audio_path) + else: + st.audio(audio_path, format="audio/mp3") + + # Show file path + st.caption(f"📁 {audio_path}") + else: + st.error("Failed to generate preview audio") + except Exception as e: + st.error(tr("tts.preview_failed", error=str(e))) + logger.exception(e) + + # ==================================================================== + # Storyboard Template Section + # ==================================================================== + + def get_template_preview_path(template_path: str, language: str = "zh_CN") -> str: + """ + Get the preview image path for a template based on language. + + Args: + template_path: Template path like "1080x1920/image_default.html" + language: Language code, either "zh_CN" or "en" + + Returns: + Path to preview image in docs/images/ + """ + # Extract size and template name from path + # e.g., "1080x1920/image_default.html" -> size="1080x1920", name="image_default" + path_parts = template_path.split('/') + if len(path_parts) >= 2: + size = path_parts[0] # e.g., "1080x1920" + template_file = path_parts[1] # e.g., "image_default.html" + template_name = template_file.replace('.html', '') # e.g., "image_default" + + # Build preview image path + # Format: docs/images/{size}/{template_name}.jpg or {template_name}_en.jpg + # Chinese uses Chinese preview, all other languages use English preview for better i18n + suffix = "" if language == "zh_CN" else "_en" + + # Try different image extensions + for ext in ['.jpg', '.png']: + preview_path = f"docs/images/{size}/{template_name}{suffix}{ext}" + if os.path.exists(preview_path): + return preview_path + + # Fallback: try without language suffix (for templates with only one version) + for ext in ['.jpg', '.png']: + preview_path = f"docs/images/{size}/{template_name}{ext}" + if os.path.exists(preview_path): + return preview_path + + # If no preview found, return empty string + return "" + + # ==================================================================== + # Workflow Selection Section (Auto-configured for Digital Human) + # ==================================================================== + with st.container(border=True): + st.markdown(f"**{tr('digital_human.section.workflow')}**") + + with st.expander(tr("help.feature_description"), expanded=False): + st.markdown(f"**{tr('help.what')}**") + st.markdown(tr("digital_human.workflow.what")) + + # Get available workflows + all_workflows = pixelle_video.media.list_workflows() + + # Auto-select digital human workflows (check file existence directly) + + # First workflow: digital_image (generates collage and copy) + first_workflow = "runninghub/digital_image.json" + first_workflow_path = "workflows/runninghub/digital_image.json" + first_workflow_found = os.path.exists(first_workflow_path) + + # Second workflow: digital_combination (generates final video) + second_workflow = "runninghub/digital_combination.json" + second_workflow_path = "workflows/runninghub/digital_combination.json" + second_workflow_found = os.path.exists(second_workflow_path) + + # First workflow status + if first_workflow_found: + st.success(f"✅ {tr('digital_human.workflow.first_step')}") + else: + st.error(f"❌ {tr('digital_human.workflow.first_step')} {tr('digital_human.workflow.not_found')}") + + # Second workflow status + if second_workflow_found: + st.success(f"✅ {tr('digital_human.workflow.second_step')}") + else: + st.error(f"❌ {tr('digital_human.workflow.second_step')} {tr('digital_human.workflow.not_found')}") + + # Show workflow info + if first_workflow_found and second_workflow_found: + st.info(f"🎯 {tr('digital_human.workflow.ready')}") + else: + st.warning(f"⚠️ {tr('digital_human.workflow.missing')}") + + # Return all style configuration parameters (简化版本，只支持local TTS) + return { + "tts_inference_mode": "local", + "tts_voice": selected_voice, + "tts_speed": tts_speed, + "first_workflow": first_workflow, + "second_workflow": second_workflow + } + diff --git a/web/i18n/locales/en_US.json b/web/i18n/locales/en_US.json index f0a0d10..89d5933 100644 --- a/web/i18n/locales/en_US.json +++ b/web/i18n/locales/en_US.json @@ -203,20 +203,20 @@ "tts.voice_selector": "Voice Selection", "tts.speed": "Speed", "tts.speed_label": "{speed}x", - "tts.voice.zh_CN_XiaoxiaoNeural": "zh-CN-XiaoxiaoNeural", - "tts.voice.zh_CN_XiaoyiNeural": "zh-CN-XiaoyiNeural", - "tts.voice.zh_CN_YunjianNeural": "zh-CN-YunjianNeural", - "tts.voice.zh_CN_YunxiNeural": "zh-CN-YunxiNeural", - "tts.voice.zh_CN_YunyangNeural": "zh-CN-YunyangNeural", - "tts.voice.zh_CN_YunyeNeural": "zh-CN-YunyeNeural", - "tts.voice.zh_CN_YunfengNeural": "zh-CN-YunfengNeural", - "tts.voice.zh_CN_liaoning_XiaobeiNeural": "zh-CN-liaoning-XiaobeiNeural", - "tts.voice.en_US_AriaNeural": "en-US-AriaNeural", - "tts.voice.en_US_JennyNeural": "en-US-JennyNeural", - "tts.voice.en_US_GuyNeural": "en-US-GuyNeural", - "tts.voice.en_US_DavisNeural": "en-US-DavisNeural", - "tts.voice.en_GB_SoniaNeural": "en-GB-SoniaNeural", - "tts.voice.en_GB_RyanNeural": "en-GB-RyanNeural", + "tts.voice.ko-KR-InJoonNeural": "ko-KR-InJoonNeural", + "tts.voice.ko-KR-SunHiNeural": "ko-KR-SunHiNeural", + "tts.voice.fr-FR-EloiseNeural": "fr-FR-EloiseNeural", + "tts.voice.fr-FR-HenriNeural": "fr-FR-HenriNeural", + "tts.voice.pt-PT-DuarteNeural": "pt-PT-DuarteNeural", + "tts.voice.pt-PT-RaquelNeural": "pt-PT-RaquelNeural", + "tts.voice.de-DE-AmalaNeural": "de-DE-AmalaNeural", + "tts.voice.de-DE-ConradNeural": "de-DE-ConradNeural", + "tts.voice.ru-RU-DmitryNeural": "ru-RU-DmitryNeural", + "tts.voice.ru-RU-SvetlanaNeural": "ru-RU-SvetlanaNeural", + "tts.voice.tr-TR-AhmetNeural": "tr-TR-AhmetNeural", + "tts.voice.tr-TR-EmelNeural": "tr-TR-EmelNeural", + "tts.voice.es-ES-AlvaroNeural": "es-ES-AlvaroNeural", + "tts.voice.es-ES-ElviraNeural": "es-ES-ElviraNeural", "tts.selector": "Workflow Selection", "tts.what": "Converts narration text to natural human-like speech (some workflows support reference audio for voice cloning)", "tts.how": "Place the exported tts_xxx.json workflow file(API format) into the workflows/selfhost/ folder (for local ComfyUI) or the workflows/runninghub/ folder (for cloud)", @@ -348,6 +348,42 @@ "pipeline.quick_create.description": "Input an idea, AI completes the entire video for you", "pipeline.custom_media.name": "Custom Media", "pipeline.custom_media.description": "Use your own photos/videos, AI adds narration and voiceover", + "pipeline.digital_human.name": "Digital Human Generate", + "pipeline.digital_human.description": "Using a piece of text, two images, and an audio clip, AI can help you generate a digital human video.", + "digital_human.section.character_assets": "😊 Character Image Upload", + "digital_human.assets.character_what": "Upload digital human character images", + "digital_human.assets.how": "Supports JPG/PNG/WebP format images, recommend clear and relevant materials", + "digital_human.assets.upload": "Upload Assets", + "digital_human.assets.upload_help": "Only supports single image upload", + "digital_human.assets.character_sucess": "Character image uploaded successfully", + "digital_human.assets.preview": "📷 Asset Preview", + "digital_human.assets.character_empty_hint": "💡 Please upload a digital human character image", + "digital_human.section.goods_assets": "👝 Product Image Upload", + "digital_human.assets.goods_what": "Upload product images", + "digital_human.assets.goods_sucess": "Product image uploaded successfully", + "digital_human.assets.goods_empty_hint": "💡 Please upload a product image", + "digital_human.section.goods_info": "🗃️ Product Information", + "digital_human.goods_title": "Product Name", + "digital_human.goods_title_placeholder": "e.g.: lipstick, coffee machine, robot vacuum, etc.", + "digital_human.goods_title_help": "Product name, cannot be empty", + "digital_human.section.workflow": "⚙️ Workflow Configuration", + "digital_human.workflow.what": "Configure the two-step workflow for digital human video generation: first step generates collage and copy, second step generates final video", + "digital_human.workflow.how": "Select appropriate workflows to handle different generation steps, ensuring output quality", + "digital_human.workflow.first_step": "First Step Workflow", + "digital_human.workflow.first_select": "Select First Step Workflow", + "digital_human.workflow.first_help": "Used to generate collage and copy based on character image, product image and product name", + "digital_human.workflow.second_step": "Second Step Workflow", + "digital_human.workflow.second_select": "Select Second Step Workflow", + "digital_human.workflow.second_help": "Used to generate final digital human video based on generated image and audio", + "digital_human.workflow.selected_info": "Selected workflow configuration", + "digital_human.workflow.auto_selected": "Auto-selected workflow", + "digital_human.workflow.configuration": "Workflow Configuration Status", + "digital_human.workflow.not_found": "not found", + "digital_human.workflow.ready": "Digital human workflows are ready, you can start generating videos", + "digital_human.workflow.missing": "Missing required digital human workflows, please check configuration", + "digital_human.workflow.advanced_options": "Advanced Options (Manual Override)", + "digital_human.workflow.manual_override": "Manual Workflow Selection (use only when needed)", + "digital_tts.what": "Convert narration text to natural human-like speech", "asset_based.section.assets": "📦 Asset Upload", "asset_based.section.video_info": "📝 Video Information", "asset_based.section.source": "⚙️ Service Configuration", diff --git a/web/i18n/locales/zh_CN.json b/web/i18n/locales/zh_CN.json index da41472..063b092 100644 --- a/web/i18n/locales/zh_CN.json +++ b/web/i18n/locales/zh_CN.json @@ -348,6 +348,8 @@ "pipeline.quick_create.description": "输入一个想法,AI帮你完成整条视频", "pipeline.custom_media.name": "自定义素材", "pipeline.custom_media.description": "用你自己的照片/视频,AI帮你配文案和配音", + "pipeline.digital_human.name": "数字人生成", + "pipeline.digital_human.description": "用一段文本、两张图片、一段音频,AI帮你生成一个数字人视频", "asset_based.section.assets": "📦 素材上传", "asset_based.section.video_info": "📝 视频信息", "asset_based.section.source": "⚙️ 服务配置", @@ -385,6 +387,30 @@ "asset_based.progress.generating_script": "📝 正在生成视频脚本...", "asset_based.progress.script_complete": "✅ 脚本生成完成", "asset_based.progress.concat_complete": "✅ 视频合成完成", + "digital_human.section.character_assets": "😊 人物形象上传", + "digital_human.assets.character_what": "上传数字人形象的图片", + "digital_human.assets.how": "支持 JPG/PNG/WebP 等格式图片，建议素材清晰且内容相关", + "digital_human.assets.upload": "上传素材", + "digital_human.assets.upload_help": "仅支持单张图片上传", + "digital_human.assets.character_sucess": "上传形象图片成功", + "digital_human.assets.preview": "📷 素材预览", + "digital_human.assets.character_empty_hint": "💡 请上传一个数字人形象图片素材", + "digital_human.section.goods_assets": "👝 商品图片上传", + "digital_human.assets.goods_what": "上传商品的图片", + "digital_human.assets.goods_sucess": "上传商品图片成功", + "digital_human.assets.goods_empty_hint": "💡 请上传一个商品图片素材", + "digital_human.section.goods_info": "🗃️ 商品信息", + "digital_human.goods_title": "商品名称", + "digital_human.goods_title_placeholder": "例如：口红、咖啡机、扫地机器人等", + "digital_human.goods_title_help": "商品的名称，不可为空", + "digital_human.section.workflow": "⚙️ 工作流加载", + "digital_human.workflow.what": "配置数字人视频生成的两步工作流：第一步生成拼图和文案，第二步生成最终视频", + "digital_human.workflow.first_step": "口播搞&拼图工作流成功加载", + "digital_human.workflow.second_step": "数字人合成工作流成功加载", + "digital_human.workflow.ready": "数字人工作流已就绪，可以开始生成视频", + "digital_human.workflow.missing": "缺少必要的数字人工作流，请检查配置", + "digital_tts.what": "将旁白文本转换为真人般的自然语音", + "faq.expand_to_view": "常见问题", "faq.load_error": "无法加载常见问题内容", "faq.more_help": "需要更多帮助？" diff --git a/web/pipelines/__init__.py b/web/pipelines/__init__.py index 1c5efa7..49859bf 100644 --- a/web/pipelines/__init__.py +++ b/web/pipelines/__init__.py @@ -26,6 +26,7 @@ from web.pipelines.base import ( # Import all pipeline UI modules to ensure they register themselves from web.pipelines import standard from web.pipelines import asset_based +from web.pipelines import digital_human __all__ = [ "PipelineUI", diff --git a/web/pipelines/digital_human.py b/web/pipelines/digital_human.py new file mode 100644 index 0000000..d2b8765 --- /dev/null +++ b/web/pipelines/digital_human.py @@ -0,0 +1,644 @@ +import os +import time +import uuid +from pathlib import Path +from typing import Any + +import streamlit as st +from loguru import logger +import httpx + +from web.i18n import tr, get_language +from web.pipelines.base import PipelineUI, register_pipeline_ui +from web.components.content_input import render_bgm_section, render_version_info +from web.components.digital_tts_config import render_style_config +from web.utils.async_helpers import run_async +from pixelle_video.config import config_manager +from pixelle_video.utils.os_util import create_task_output_dir + +class DigitalHumanPipelineUI(PipelineUI): + """ + UI for the Digital_Human Video Generation Pipeline. + Generates videos from user-provided assets (images&videos&audio). + """ + name = "digital_human" + icon = "🤖" + + @property + def display_name(self): + return tr("pipeline.digital_human.name") + + @property + def description(self): + return tr("pipeline.digital_human.description") + + def render(self, pixelle_video: Any): + # Three-column layout + left_col, middle_col, right_col = st.columns([1, 1, 1]) + + # ==================================================================== + # Left Column: Asset Upload + # ==================================================================== + with left_col: + asset_params = self.render_digital_human_input() + # bgm_params = render_bgm_section(key_prefix="asset_") + render_version_info() + + # ==================================================================== + # Middle Column: Video Configuration + # ==================================================================== + with middle_col: + # Style configuration (TTS, template, workflow, etc.) + style_params = render_style_config(pixelle_video) + + # # ==================================================================== + # # Right Column: Output Preview + # # ==================================================================== + with right_col: + # Combine all parameters + video_params = { + **asset_params, + **style_params + } + + self._render_output_preview(pixelle_video, video_params) + + def render_digital_human_input(self) -> dict: + """Render digital human character image upload section""" + with st.container(border=True): + st.markdown(f"**{tr('digital_human.section.character_assets')}**") + + with st.expander(tr("help.feature_description"), expanded=False): + st.markdown(f"**{tr('help.what')}**") + st.markdown(tr("digital_human.assets.character_what")) + st.markdown(f"**{tr('help.how')}**") + st.markdown(tr("digital_human.assets.how")) + + # File uploader for multiple files + uploaded_files = st.file_uploader( + tr("digital_human.assets.upload"), + type=["jpg", "jpeg", "png", "webp"], + accept_multiple_files=True, + help=tr("digital_human.assets.upload_help"), + key="character_files" + ) + + # Save uploaded files to temp directory with unique session ID + character_asset_paths = [] + if uploaded_files: + import uuid + session_id = str(uuid.uuid4()).replace('-', '')[:12] + temp_dir = Path(f"temp/assets_{session_id}") + temp_dir.mkdir(parents=True, exist_ok=True) + + for uploaded_file in uploaded_files: + file_path = temp_dir / uploaded_file.name + with open(file_path, "wb") as f: + f.write(uploaded_file.getbuffer()) + character_asset_paths.append(str(file_path.absolute())) + + st.success(tr("digital_human.assets.character_sucess")) + + # Preview uploaded assets + with st.expander(tr("digital_human.assets.preview"), expanded=True): + # Show in a grid (3 columns) + cols = st.columns(3) + for i, (file, path) in enumerate(zip(uploaded_files, character_asset_paths)): + with cols[i % 3]: + # Check if image + ext = Path(path).suffix.lower() + if ext in [".jpg", ".jpeg", ".png", ".webp"]: + st.image(file, caption=file.name, use_container_width=True) + else: + st.info(tr("digital_human.assets.character_empty_hint")) + + with st.container(border=True): + st.markdown(f"**{tr('digital_human.section.goods_assets')}**") + + with st.expander(tr("help.feature_description"), expanded=False): + st.markdown(f"**{tr('help.what')}**") + st.markdown(tr("digital_human.assets.goods_what")) + st.markdown(f"**{tr('help.how')}**") + st.markdown(tr("digital_human.assets.how")) + + # File uploader for multiple files + uploaded_files = st.file_uploader( + tr("digital_human.assets.upload"), + type=["jpg", "jpeg", "png", "webp"], + accept_multiple_files=True, + help=tr("digital_human.assets.upload_help"), + key="goods_files" + ) + + # Save uploaded files to temp directory with unique session ID + goods_asset_paths = [] + if uploaded_files: + import uuid + session_id = str(uuid.uuid4()).replace('-', '')[:12] + temp_dir = Path(f"temp/assets_{session_id}") + temp_dir.mkdir(parents=True, exist_ok=True) + + for uploaded_file in uploaded_files: + file_path = temp_dir / uploaded_file.name + with open(file_path, "wb") as f: + f.write(uploaded_file.getbuffer()) + goods_asset_paths.append(str(file_path.absolute())) + + st.success(tr("digital_human.assets.goods_sucess")) + + # Preview uploaded assets + with st.expander(tr("digital_human.assets.preview"), expanded=True): + # Show in a grid (3 columns) + cols = st.columns(3) + for i, (file, path) in enumerate(zip(uploaded_files, goods_asset_paths)): + with cols[i % 3]: + # Check if image + ext = Path(path).suffix.lower() + if ext in [".jpg", ".jpeg", ".png", ".webp"]: + st.image(file, caption=file.name, use_container_width=True) + else: + st.info(tr("digital_human.assets.goods_empty_hint")) + + # Goods Name or Description + with st.container(border=True): + st.markdown(f"**{tr('digital_human.section.goods_info')}**") + + goods_title = st.text_input( + tr("digital_human.goods_title"), + placeholder=tr("digital_human.goods_title_placeholder"), + help=tr("digital_human.goods_title_help"), + key="goods_title" + ) + + return { + "character_assets": character_asset_paths, + "goods_title": goods_title, + "goods_assets": goods_asset_paths + } + + def _render_output_preview(self, pixelle_video: Any, video_params: dict): + """Render output preview section""" + with st.container(border=True): + st.markdown(f"**{tr('section.video_generation')}**") + + # Check configuration + if not config_manager.validate(): + st.warning(tr("settings.not_configured")) + + # Get input data + character_assets = video_params.get("character_assets", []) + goods_assets = video_params.get("goods_assets", []) + goods_title = video_params.get("goods_title", "") + first_workflow = video_params.get("first_workflow") + second_workflow = video_params.get("second_workflow") + tts_voice = video_params.get("tts_voice", "zh-CN-YunjianNeural") + tts_speed = video_params.get("tts_speed", 1.2) + + # Debug: 显示获取到的TTS参数 + logger.info(f"🔧 获取到的TTS参数:") + logger.info(f" - tts_voice: {tts_voice}") + logger.info(f" - tts_speed: {tts_speed}") + logger.info(f" - video_params中的tts_voice: {video_params.get('tts_voice', 'NOT_FOUND')}") + logger.info(f" - 所有video_params: {video_params}") + + # Validation + if not character_assets or not goods_assets: + st.info("请上传人物形象图片和商品图片") + st.button( + tr("btn.generate"), + type="primary", + use_container_width=True, + disabled=True, + key="digital_human_generate_disabled" + ) + return + + if not goods_title: + st.warning("请输入商品名称") + st.button( + tr("btn.generate"), + type="primary", + use_container_width=True, + disabled=True, + key="digital_human_generate_disabled_no_title" + ) + return + + # Workflow validation + if not first_workflow or not second_workflow: + st.warning("请选择工作流配置") + st.button( + tr("btn.generate"), + type="primary", + use_container_width=True, + disabled=True, + key="digital_human_generate_disabled_no_workflow" + ) + return + + # Generate button + if st.button(tr("btn.generate"), type="primary", use_container_width=True, key="digital_human_generate"): + # Validate + if not config_manager.validate(): + st.error(tr("settings.not_configured")) + st.stop() + + # Show progress + progress_bar = st.progress(0) + status_text = st.empty() + + start_time = time.time() + + try: + # Define async generation function + async def generate_digital_human_video(): + # Create task directory + task_dir, task_id = create_task_output_dir() + logger.info(f"📁 Task directory created: {task_dir}") + + # ============================================================ + # Step 1: Call first workflow (character_image + goods_image + goods_title) + # ============================================================ + status_text.text("步骤 1/3: 调用第一步工作流生成拼图和文案...") + progress_bar.progress(10) + + # Directly load first workflow file + import json + from pathlib import Path + + first_workflow_path = Path("workflows/runninghub/digital_image.json") + if not first_workflow_path.exists(): + raise Exception(f"第一步工作流文件不存在: {first_workflow_path}") + + with open(first_workflow_path, 'r', encoding='utf-8') as f: + first_workflow_config = json.load(f) + + # Build workflow parameters for first workflow + # 注意：参数名需要根据你的 workflow 实际参数来调整 + # 尝试多种可能的参数名 + first_workflow_params = { + # 根据之前的日志尝试的参数名 + "firstimage": character_assets[0], + "secondimage": goods_assets[0], + "goodstype": goods_title + } + + logger.info(f"🔧 传递给第一步工作流的参数:") + for key, value in first_workflow_params.items(): + if isinstance(value, str) and len(value) > 100: + logger.info(f" - {key}: {value[:100]}... (文件路径)") + else: + logger.info(f" - {key}: {value}") + + # Execute first workflow + kit = await pixelle_video._get_or_create_comfykit() + + # Use workflow_id for runninghub workflows + if first_workflow_config.get("source") == "runninghub" and "workflow_id" in first_workflow_config: + workflow_input = first_workflow_config["workflow_id"] + else: + workflow_input = str(first_workflow_path) + + logger.info(f"执行第一步工作流: {workflow_input}") + first_result = await kit.execute(workflow_input, first_workflow_params) + + if first_result.status != "completed": + raise Exception(f"第一步工作流执行失败: {first_result.msg}") + + # Debug: Log the actual result structure + logger.info(f"🔍 第一步工作流返回结果调试:") + logger.info(f"- result type: {type(first_result)}") + logger.info(f"- result.__dict__: {first_result.__dict__}") + + if hasattr(first_result, 'images'): + logger.info(f"- images: {first_result.images}") + if hasattr(first_result, 'texts'): + logger.info(f"- texts: {first_result.texts}") + if hasattr(first_result, 'videos'): + logger.info(f"- videos: {first_result.videos}") + if hasattr(first_result, 'outputs'): + logger.info(f"- outputs keys: {list(first_result.outputs.keys()) if first_result.outputs else 'None'}") + if first_result.outputs: + for node_id, node_output in first_result.outputs.items(): + logger.info(f" - 节点 {node_id}: {node_output}") + + # 尝试获取所有可能的属性 + for attr in dir(first_result): + if not attr.startswith('_'): + try: + value = getattr(first_result, attr) + if not callable(value): + logger.info(f"- {attr}: {value}") + except: + pass + + # Extract results from first workflow + generated_image_url = None + generated_text = None + generated_video_url = None + + # 检查是否直接返回了视频（一步到位的工作流） + if hasattr(first_result, 'videos') and first_result.videos: + generated_video_url = first_result.videos[0] + logger.info(f"✅ 第一步工作流直接生成了视频: {generated_video_url}") + + # 如果直接生成了视频，跳过后续步骤 + progress_bar.progress(100) + status_text.text(tr("status.success")) + + # Download video to local + final_video_path = os.path.join(task_dir, "final.mp4") + timeout = httpx.Timeout(300.0) # 简化timeout设置 + async with httpx.AsyncClient(timeout=timeout) as client: + response = await client.get(generated_video_url) + response.raise_for_status() + with open(final_video_path, 'wb') as f: + f.write(response.content) + + return final_video_path + + # 如果没有直接返回视频，尝试提取图片和文本（原来的逻辑） + # Extract image - try direct access first + if hasattr(first_result, 'images') and first_result.images: + generated_image_url = first_result.images[0] + logger.info(f"✅ 从 result.images 获取图片: {generated_image_url}") + elif hasattr(first_result, 'outputs') and first_result.outputs: + # Try to find image in outputs + for node_id, node_output in first_result.outputs.items(): + logger.info(f"- 检查节点 {node_id}: {list(node_output.keys()) if isinstance(node_output, dict) else type(node_output)}") + if isinstance(node_output, dict) and 'images' in node_output: + images = node_output['images'] + if images and len(images) > 0: + generated_image_url = images[0] + logger.info(f"✅ 从 outputs[{node_id}].images 获取图片: {generated_image_url}") + break + + # Extract text - try direct access first + if hasattr(first_result, 'texts') and first_result.texts: + generated_text = first_result.texts[0] + logger.info(f"✅ 从 result.texts 获取文本: {generated_text[:50]}...") + elif hasattr(first_result, 'outputs') and first_result.outputs: + # Try to find text in outputs + for node_id, node_output in first_result.outputs.items(): + if isinstance(node_output, dict) and 'text' in node_output: + text_list = node_output['text'] + if text_list and len(text_list) > 0: + generated_text = text_list[0] + logger.info(f"✅ 从 outputs[{node_id}].text 获取文本: {generated_text[:50]}...") + break + + if not generated_image_url: + # 如果没有找到图片，尝试查看是否有其他类型的输出 + logger.error("❌ 第一步工作流未返回图片，尝试查找其他输出...") + if hasattr(first_result, 'outputs') and first_result.outputs: + for node_id, node_output in first_result.outputs.items(): + logger.info(f"- 节点 {node_id} 输出: {node_output}") + + # 尝试使用第一个可用的图片输出（如果有的话） + if hasattr(first_result, 'outputs') and first_result.outputs: + for node_id, node_output in first_result.outputs.items(): + if isinstance(node_output, dict): + # 检查所有可能的图片字段 + for key in ['images', 'image', 'output_image', 'result_image']: + if key in node_output and node_output[key]: + if isinstance(node_output[key], list) and len(node_output[key]) > 0: + generated_image_url = node_output[key][0] + logger.info(f"✅ 从 outputs[{node_id}].{key} 获取图片: {generated_image_url}") + break + elif isinstance(node_output[key], str): + generated_image_url = node_output[key] + logger.info(f"✅ 从 outputs[{node_id}].{key} 获取图片: {generated_image_url}") + break + if generated_image_url: + break + + if not generated_image_url: + raise Exception("第一步工作流未返回图片，请检查工作流配置") + + if not generated_text: + # 如果没有找到文本，尝试查找其他文本输出 + logger.warning("⚠️ 第一步工作流未返回文本，尝试查找其他文本输出...") + if hasattr(first_result, 'outputs') and first_result.outputs: + for node_id, node_output in first_result.outputs.items(): + if isinstance(node_output, dict): + # 检查所有可能的文本字段 + for key in ['text', 'texts', 'output_text', 'result_text', 'description', 'caption']: + if key in node_output and node_output[key]: + if isinstance(node_output[key], list) and len(node_output[key]) > 0: + generated_text = node_output[key][0] + logger.info(f"✅ 从 outputs[{node_id}].{key} 获取文本: {generated_text[:50]}...") + break + elif isinstance(node_output[key], str): + generated_text = node_output[key] + logger.info(f"✅ 从 outputs[{node_id}].{key} 获取文本: {generated_text[:50]}...") + break + if generated_text: + break + + # 如果还是没有文本，使用默认文本 + if not generated_text: + generated_text = f"为您推荐这款{goods_title}，品质优良，值得拥有！" + logger.warning(f"⚠️ 使用默认文本: {generated_text}") + + logger.info(f"✅ 第一步完成: 图片={generated_image_url}, 文本={generated_text[:50]}...") + + # Download generated image to local + generated_image_path = os.path.join(task_dir, "generated_image.png") + timeout = httpx.Timeout(10.0) # 简化timeout设置 + async with httpx.AsyncClient(timeout=timeout) as client: + response = await client.get(generated_image_url) + response.raise_for_status() + with open(generated_image_path, 'wb') as f: + f.write(response.content) + + progress_bar.progress(40) + status_text.text("步骤 2/3: 将文案转换为语音...") + + # ============================================================ + # Step 2: TTS conversion (text -> audio) + # ============================================================ + audio_path = os.path.join(task_dir, "narration.mp3") + await pixelle_video.tts( + text=generated_text, + output_path=audio_path, + inference_mode="local", # 明确指定推理模式 + voice=tts_voice, # 使用voice而不是voice_id + speed=tts_speed + ) + + logger.info(f"✅ 第二步完成: 语音={audio_path}") + progress_bar.progress(70) + status_text.text("步骤 3/3: 调用第二步工作流生成最终视频...") + + # ============================================================ + # Step 3: Call second workflow (generated_image + audio) + # ============================================================ + # Directly load second workflow file + second_workflow_path = Path("workflows/runninghub/digital_combination.json") + if not second_workflow_path.exists(): + raise Exception(f"第二步工作流文件不存在: {second_workflow_path}") + + with open(second_workflow_path, 'r', encoding='utf-8') as f: + second_workflow_config = json.load(f) + + # Build workflow parameters for second workflow + # 注意：参数名需要根据你的 workflow 实际参数来调整 + # 尝试多种可能的参数名 + second_workflow_params = { + # 尝试常见的参数名 + "videoimage": generated_image_path, # 生成的图片路径 + "audio": audio_path, # 音频路径 + } + + logger.info(f"🔧 传递给第二步工作流的参数:") + for key, value in second_workflow_params.items(): + logger.info(f" - {key}: {value}") + + # Execute second workflow + if second_workflow_config.get("source") == "runninghub" and "workflow_id" in second_workflow_config: + workflow_input = second_workflow_config["workflow_id"] + else: + workflow_input = str(second_workflow_path) + + logger.info(f"执行第二步工作流: {workflow_input}") + second_result = await kit.execute(workflow_input, second_workflow_params) + + if second_result.status != "completed": + raise Exception(f"第二步工作流执行失败: {second_result.msg}") + + # Debug: Log the second workflow result structure + logger.info(f"🔍 第二步工作流返回结果调试:") + logger.info(f"- result type: {type(second_result)}") + logger.info(f"- result.__dict__: {second_result.__dict__}") + + if hasattr(second_result, 'videos'): + logger.info(f"- videos: {second_result.videos}") + if hasattr(second_result, 'images'): + logger.info(f"- images: {second_result.images}") + if hasattr(second_result, 'texts'): + logger.info(f"- texts: {second_result.texts}") + if hasattr(second_result, 'outputs'): + logger.info(f"- outputs keys: {list(second_result.outputs.keys()) if second_result.outputs else 'None'}") + if second_result.outputs: + for node_id, node_output in second_result.outputs.items(): + logger.info(f" - 节点 {node_id}: {node_output}") + + # 尝试获取所有可能的属性 + for attr in dir(second_result): + if not attr.startswith('_'): + try: + value = getattr(second_result, attr) + if not callable(value): + logger.info(f"- {attr}: {value}") + except: + pass + + # Extract video from second workflow result + generated_video_url = None + + if hasattr(second_result, 'videos') and second_result.videos: + generated_video_url = second_result.videos[0] + logger.info(f"✅ 从 result.videos 获取视频: {generated_video_url}") + elif hasattr(second_result, 'outputs') and second_result.outputs: + for node_id, node_output in second_result.outputs.items(): + logger.info(f"- 检查节点 {node_id}: {list(node_output.keys()) if isinstance(node_output, dict) else type(node_output)}") + if isinstance(node_output, dict) and 'videos' in node_output: + videos = node_output['videos'] + if videos and len(videos) > 0: + generated_video_url = videos[0] + logger.info(f"✅ 从 outputs[{node_id}].videos 获取视频: {generated_video_url}") + break + + if not generated_video_url: + # 如果没有找到视频，尝试查看是否有其他类型的输出 + logger.error("❌ 第二步工作流未返回视频，尝试查找其他输出...") + if hasattr(second_result, 'outputs') and second_result.outputs: + for node_id, node_output in second_result.outputs.items(): + logger.info(f"- 节点 {node_id} 输出: {node_output}") + + # 尝试使用第一个可用的视频输出（如果有的话） + if hasattr(second_result, 'outputs') and second_result.outputs: + for node_id, node_output in second_result.outputs.items(): + if isinstance(node_output, dict): + # 检查所有可能的视频字段 + for key in ['videos', 'video', 'output_video', 'result_video', 'mp4']: + if key in node_output and node_output[key]: + if isinstance(node_output[key], list) and len(node_output[key]) > 0: + generated_video_url = node_output[key][0] + logger.info(f"✅ 从 outputs[{node_id}].{key} 获取视频: {generated_video_url}") + break + elif isinstance(node_output[key], str): + generated_video_url = node_output[key] + logger.info(f"✅ 从 outputs[{node_id}].{key} 获取视频: {generated_video_url}") + break + if generated_video_url: + break + + if not generated_video_url: + raise Exception("第二步工作流未返回视频，请检查工作流配置") + + logger.info(f"✅ 第三步完成: 视频={generated_video_url}") + + # Download video to local + final_video_path = os.path.join(task_dir, "final.mp4") + timeout = httpx.Timeout(300.0) # 简化timeout设置 + async with httpx.AsyncClient(timeout=timeout) as client: + response = await client.get(generated_video_url) + response.raise_for_status() + with open(final_video_path, 'wb') as f: + f.write(response.content) + + return final_video_path + + # Execute async generation + final_video_path = run_async(generate_digital_human_video()) + + total_time = time.time() - start_time + progress_bar.progress(100) + status_text.text(tr("status.success")) + + # Display result + st.success(tr("status.video_generated", path=final_video_path)) + + st.markdown("---") + + # Video info + if os.path.exists(final_video_path): + file_size_mb = os.path.getsize(final_video_path) / (1024 * 1024) + + info_text = ( + f"⏱️ {tr('info.generation_time')} {total_time:.1f}s " + f"📦 {file_size_mb:.2f}MB" + ) + st.caption(info_text) + + st.markdown("---") + + # Video preview + st.video(final_video_path) + + # Download button + with open(final_video_path, "rb") as video_file: + video_bytes = video_file.read() + video_filename = os.path.basename(final_video_path) + st.download_button( + label="⬇️ 下载视频" if get_language() == "zh_CN" else "⬇️ Download Video", + data=video_bytes, + file_name=video_filename, + mime="video/mp4", + use_container_width=True + ) + else: + st.error(tr("status.video_not_found", path=final_video_path)) + + except Exception as e: + status_text.text("") + progress_bar.empty() + st.error(tr("status.error", error=str(e))) + logger.exception(e) + st.stop() + + +# Register self +register_pipeline_ui(DigitalHumanPipelineUI) + diff --git a/workflows/runninghub/digital_combination.json b/workflows/runninghub/digital_combination.json new file mode 100644 index 0000000..2ea1ce2 --- /dev/null +++ b/workflows/runninghub/digital_combination.json @@ -0,0 +1,4 @@ +{ + "source": "runninghub", + "workflow_id": "2003717471859294210" +} \ No newline at end of file diff --git a/workflows/runninghub/digital_image.json b/workflows/runninghub/digital_image.json new file mode 100644 index 0000000..b04f127 --- /dev/null +++ b/workflows/runninghub/digital_image.json @@ -0,0 +1,4 @@ +{ + "source": "runninghub", + "workflow_id": "2004120336125861890" +} \ No newline at end of file